NousResearch/atropos
1
0
Fork 0
mirror of https://github.com/NousResearch/atropos.git synced 2026-04-23 16:54:56 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

vllm restart 1

Browse source
This commit is contained in:
Jai Suphavadeeprasit 2026-02-12 12:28:00 -05:00
parent 6bd0296bac
commit 9dcb362aba
2 changed files with 22 additions and 6 deletions
Download patch file Download diff file Expand all files Collapse all files

8
example_trainer/trainers.py
View file

@ -891,9 +891,15 @@ def _launch_vllm_with_lora(config: TrainingConfig, adapter_path: str) -> Optiona
print(f" Launching: {' '.join(cmd)}")
print(f" Adapter: {adapter_path}")
# Log vLLM output to file for debugging
vllm_log_path = os.path.join(config.save_path, "vllm_internal.log")
print(f" vLLM log: {vllm_log_path}")
try:
proc = subprocess.Popen(cmd, env=env)
vllm_log_file = open(vllm_log_path, "w")
proc = subprocess.Popen(cmd, env=env, stdout=vllm_log_file, stderr=subprocess.STDOUT)
print(f" vLLM PID: {proc.pid}")
print(f" NOTE: vLLM without --enforce-eager compiles CUDA graphs on startup (takes 1-3 min)...")
# Wait for server to be ready
if not wait_for_vllm_ready(config.vllm_port, timeout=180):